The use of Remote sensing as a monitoring tool for coastal defence issues in the Wadden Sea

7/31/2019 The use of Remote sensing as a monitoring tool for coastal defence issues in the Wadden Sea

1/45

The u se o f Rem ot e sens ing

a s a m o n i t o r i n g t o o l f o rcoas ta l de fence issues in t he

W adden Sea

Narangerel Davaasuren, Johan Stapel, Cor Smit,

Norbert Dankers

Report number C057/12

I MARES Wageningen URInstitute for Marine Resources & Ecosystem Studies

Client: Rijkswaterstaat Waterdienst

Deltaprogramma Wadden

Rick Hoeksema

Postbus 178200 AA Lelystad

Publication date: 27 June 2012


2/45

2 of 45 Report number C057/12

I MARES is:

an independent, objective and authoritative institute that provides knowledge necessary foran integrated sustainable protection, exploitation and spatial use of the sea and coastal

zones;

an institute that provides knowledge necessary for an integrated sustainable protection,exploitation and spatial use of the sea and coastal zones;

a key, proactive player in national and international marine networks (including ICES andEFARO).

photographs title page: images- left- Landsat TM, 1996; right- WorldView-2, 2006

Images copyright: Landsat- Global Land Cover Facility, University of Maryland, USA;

WorldView-2- Digital Globe, USA.

Copyright restrictions- WorldView-2 image received under special agreement from Digital Globe on ERDAS

IMAGINE-Digital globe 2012 Geospatial Challenge contest, images cannot be copied and reproduced without

obtaining prior permission from Digital Globe. Landsat images- acknowledgement required.

P.O. Box 68 P.O. Box 77 P.O. Box 57 P.O. Box 167

1970 AB IJmuiden 4400 AB Yerseke 1780 AB Den Helder 1790 AD Den Burg Texel

Phone: +31 (0)317 48 09 00 Phone: +31 (0)317 48 09 00 Phone: +31 (0)317 48 09 00 Phone: +31 (0)317 48 09 00

Fax: +31 (0)317 48 73 26 Fax: +31 (0)317 48 73 59 Fax: +31 (0)223 63 06 87 Fax: +31 (0)317 48 73 62

E-Mail: [email protected] E-Mail: [email protected] E-Mail: [email protected] E-Mail: [email protected]

www.imares.wur.nl www.imares.wur.nl www.imares.wur.nl www.imares.wur.nl

2012 IMARES Wageningen UR

IMARES, institute of Stichting DLO is

registered in the Dutch trade

record nr. 09098104,

BTW nr. NL 806511618

The Management of IMARES is not responsible for resulting damage,

as well as for damage resulting from the application of results or

research obtained by IMARES, its clients or any claims related to the

application of information found within its research. This report has

been made on the request of the client and is wholly the client's

property. This report may not be reproduced and/or published

partially or in its entirety without the express written consent of the

client.

A_4_3_2-V12.3


3/45

Report number C057/12 3 of 45

Samenva t t i ng

Grote delen van Nederland liggen onder zeeniveau. De Waddenzee, met de ervoor liggende eilanden, is

een belangrijke natuurlijke barrire die de noordelijke kuststrook beschermt tegen de Noordzee. De

Waddenzee is een ondiep dynamisch en troebel ecosysteem met veel bij laagwater droogvallende platen.

Het Delta Programma Waddenzee behelst de Waddenzee, de Wadden eilanden, het Eems-Dollard

estuarium en de (Wadden)kustzones van Noord Holland, Friesland en Groningen. Klimaat verandering

kan leiden tot verhoging van het zeeniveau, meer extreme weersituaties leidend tot hogere

stormvloeden, meer zoetwaterafvoer, frequentere stormen en veranderingen in de richting van de

golfaanval. Daardoor kunnen de eilanden sterker eroderen en droogvallende platen en kwelders, de

natuurlijke beschermingszones voor de vastelandsdijken, kunnen verdwijnen. In hoeverre dat zal

gebeuren en met welk tempo is nog onduidelijk.

Remote sensing kan een bruikbare en relatief goedkope methode zijn om specifieke karakteristieken van

droogvallende platen waar te nemen. Het geeft de mogelijkheid om ruimtelijke en temporele variaties in

sediment en biogene structuren te kwantificeren en te volgen. De techniek kan informatie uit metingen

op enkele specifieke locaties vertalen naar een groter gebied. In aanvulling daarop kunnen historische

beelden gebruikt worden om trends in ontwikkelingen over grote gebieden te detecteren. Er zijn beelden

beschikbaar vanaf 1975. Zodoende kan het gedrag van belangrijke ecosysteemcomponenten over een

lange periode beschreven worden

De toepassing van Remote Sensing in kustbeschermingsbeleid en beheer vergt specifieke aanpassingen

aan bestaande classificatie technieken. Het bepalen van relevante parameters in een troebel, zeer

dynamisch ondiep systeem als de Waddenzee is een grote uitdaging. De doelstellingen van het huidige

rapport, als onderdeel van de monitoring component in het Delta Programma Waddenzee, zijn het

beoordelen van mogelijkheden en limitaties van de toepassing van RS als instrument voor het monitoren

van variabelen die relevant geacht worden in het kader van kustveiligheid. Gedacht moet worden aan

sedimentkarakteristieken, schelpdierbanken en dynamisch gedrag van geulen en prielen.

De studie haakt aan bij internationale Remote Sensing expertise en methoden en richt zich op aspecten

die belangrijk zijn voor kustbeheer in de Nederlandse Waddenzee.

De vereiste resolutie, ofwel het oplossend vermogen, om sedimentkarakteristieken en kenmerken van

droogvallende schelpdierbanken te kunnen bepalen moet beter zijn dan 30 meter. De beschikbaarheid

van historische multispectrale satellietbeelden voor de Waddenzee, zonder wolken en bij laagwater, is

beperkt. Meestal zijn minder dan 4 beelden per jaar beschikbaar. Voor het bepalen van trends in sterkvariabele of hoog dynamische parameters is toepassing van Remote Sensing op basis van

satellietbeelden daarom niet aan te raden. Voor relatief stabiele structuren zoals kwelder(randen)

schelpdierbanken, geulen en prielen is RS zeer bruikbaar bij het volgen van ruimtelijke en temporele

veranderingen

Historische Landsat beelden laten verschillen zien in sediment karakteristieken zoals vochtgehalte en

zand/slib verhouding. De locaties en grootte van schelpdierbanken (mossel en oester) konden met

redelijke nauwkeurigheid gevolgd worden op de beelden bij laagwater. De ontwikkeling van geulen en

prielen in de richting van de kust kon in kaart worden gebracht waarmee mogelijk potentieel voor

kustbescherming gevaarlijke ontwikkelingen tijdig zijn te detecteren

Mogelijke toekomstige toepassingen voor het monitoren van veranderingen en toepassingen voor beleids-

en beheersondersteunende modellen wordt besproken.


4/45


S u m m a r y

Large parts of the Netherlands are below sea level. The Wadden Sea is an important natural barrier,

protecting the northern coast of the Netherlands from the North Sea. The Wadden Sea is a shallow, partly

intertidal, turbid and dynamic ecosystem. Climate change may, among others, lead to increased sea level

rise, possibly causing drowning of the Wadden Sea and losing its protective function. If and how these

changes occur and at which pace is still unclear.

Remote sensing can be a very useful and relatively inexpensive tool in detecting specific characteristics of

emerging tidal flats. Remote sensing techniques offer an alternative option to track, detect and to analyse

spatial and temporal variations in e.g. sediment characteristics and biogenic structures. The technique

can be used to translate information on habitat characteristics from point measurements to a wider

geographical area. In addition, Remote sensing databases of historic images, going back to 1975, offerthe possibility for trend analyses of large areas as to provide valuable information on the behaviour of

important components of the Wadden Sea area in the past.

The application of remote sensing in coastal defence policy and management requires very specific

adaptations to existing classification techniques. Detection of relevant parameters from a turbid, highly

dynamic shallow ecosystem as the Wadden Sea is a challenging task. The objectives of this report as part

of the monitoring component in the Delta Programme Wadden Sea, is to assess the possibilities and

limitations of the application of Remote sensing as a tool for monitoring parameters that are relevant to

coastal defence objectives, i.e. sediment qualities, mussel and oyster beds and channel and gully

dynamics. The study provides links to international Remote sensing expertise and methods. It also

focusses on tools for testing Remote sensing applicability for answering questions related to coastal

defence issues in the Dutch Wadden Sea.

The required resolution for detecting sediment qualities and mussel and oyster beds is 30 m or higher.

The number of these medium resolution historical multi-spectral satellite images for the Wadden Sea is

limited (up to four images per year). Satellite based Remote sensing of historical trends of highly

dynamic and variable parameters are therefore not recommended. For structures like salt marshes,

mussel and oyster beds, intertidal flats and gullies and channels, however, Remote sensing is highly

applicable for measuring temporal or spatial changes.

Historical Landsat data showed different characteristics of intertidal sediments in terms of water content

and sand and mud content. The locations of mussel and oyster beds can be fairly well detected on images

acquired during low tide. Time series of satellite images showed progression of gullies and channelstowards the coast, potentially endangering defence structures.

Future perspectives of using remote sensing for monitoring changes in the Wadden Sea and application in

scenario and management support models are discussed.


5/45


Conten ts

Samenvatting ............................................................................................................ 3Summary ................................................................................................................. 4Contents ................................................................................................................... 51 Introduction ..................................................................................................... 7

1.1 Delta Programme Wadden Sea .................................................................. 71.2 Monitoring .............................................................................................. 71.3 Remote Sensing ...................................................................................... 8

1.3.1 Classification of sediments using satellite data .................................. 81.3.2 Identification and locations of mussel and oyster beds ........................ 81.3.3 Detection of temporal changes in gullies and channels ........................ 9

1.4 The research objectives of the study ........................................................... 91.4.1 The general objective .................................................................... 91.4.2 The specific objective .................................................................... 9

1.5 The limitations of the study ....................................................................... 91.5.1 Research methodology ................................................................ 101.5.2 Structure of the report ................................................................. 10

2 Assignment .................................................................................................... 113 Materials and Methods ..................................................................................... 12

3.1 Historical medium resolution satellite data ................................................. 123.1.1 Landsat ..................................................................................... 133.1.2 ASTER ....................................................................................... 133.1.3 ERS 14

3.2 Application of Remote sensing in the Wadden Sea ...................................... 153.3 Classification ......................................................................................... 16

3.3.1 Classification of sediments ........................................................... 163.3.2 Classification of mussel and oyster beds ......................................... 163.3.3 Temporal change detection in gullies and channels .......................... 17

3.4 Selection of Remote sensing classification method ...................................... 173.5 Selection of pilot parameters and pilot area ............................................... 183.6 Factors restricting satellite data availability- weather and tides ..................... 193.7 Inventory of satellite images ................................................................... 203.8 Pre-processing ...................................................................................... 20

3.8.1 Atmospheric correction of Landsat images ...................................... 203.8.2 Spectral analysis ......................................................................... 213.8.3 Result from Modified Soil Adjusted Vegetation index ........................ 273.8.4 Radar images ............................................................................. 28

4 Results .......................................................................................................... 294.1 Classification ......................................................................................... 294.1.1 Sediment ................................................................................... 29

4.1.2 Classification of mussel/oyster beds ............................................... 30


6/45


4.1.3 Temporal development of gullies and channels ................................ 315 Future applications .......................................................................................... 33

5.1 Storm database ..................................................................................... 335.2 Integrated model ................................................................................... 335.3 High resolution satellite data ................................................................... 34

6 Discussion ..................................................................................................... 37References .............................................................................................................. 39Quality Assurance .................................................................................................... 41Justification ............................................................................................................. 42Appendix 1 Complete overview of available cloud-free and low-tide Landsat satellite images43Appendix 2- Glossary of Technical Terms ..................................................................... 44


7/45


1 I n t r o d u c t i o n

1 .1 Del ta Prog ram m e Wadden SeaThe Delta Programme Wadden Sea area addresses the Wadden Sea, the Wadden Sea islands, the Eems-

Dollard estuary and the coastal zones of the provinces of North Holland, Friesland and Groningen that

border the Wadden Sea and the Eems-Dollard estuary. Climate change may lead to increased sea level

rise and a change to more extremes in the tidal amplitude, the atmospheric precipitation pattern and in

the occurrence and timing of storm events and the angle of wave attack relative to the coastline. As a

consequence it is expected that the North Sea side of the Wadden Sea island will experience other and

increased coastal erosion and the intertidal mudflats and salt marshes in the Wadden Sea Area, the

natural coastal defence of the mainland, could drown. If and how these changes occur and at which pace

is still unclear.

The Delta Programme Wadden Sea Area aims:

To develop an integrated routine which should secure the safety of the islands and main lands

coasts. The objective is to integrate water safety with other functions of the Wadden Sea Area, which

are: nature, recreation and tourism, and sustainable economic activities.

To monitor the effects of climate change on ecosystems and developments in the Wadden Sea

Area.

For appropriate policy development it is important to know:

If and which system changes are likely to occur in response to climate change and at what pace,

If these changes negatively affect safety or natural and socioeconomic values,

If these changes are disagreeable from a public opinions point of view, and If moments in time can be signposted at which adaptive measures should be taken or beyond

which changes are irreversible and other less appealing or expensive measures are required to meet at

least the flood defence safety norm (adaptation tipping points).

1 .2 Mon i to r ingThe objectives of a Monitoring programme as one of the areas of attention in the Delta Programme

Wadden Sea Area are (Stapel and Dankers, 2011):

To create coherence between and optimisation of different monitoring activities and data

information management (national and international), To integrate the long term measurements suggested from other areas of attention,

To interpret data in order to answer the policy questions raised above and to contribute to the

objectives of the Delta Programme Wadden Sea Area as a whole

The aim of monitoring is to detect and to monitor the effects of climate change on ecosystems and

developments in the Wadden Sea area, which could have an impact on the coastal integrity.

Remote sensing is a cost-effective and time saving method to obtain measurements from large areas

(altitude, sediment composition, channel and gully morphology, geological and biogenic structures,

zonation and vegetation characteristics). Remote sensing techniques offer an alternative option to track,

detect and analyse spatial and temporal variations in e.g. sediment characteristics and biogenicstructures. The technique may be used to translate information on habitat characteristics from point


8/45


measurements to a wider geographical area. In addition Remote sensing databases of historic images,

going back to 1975, offer the possibility for trend analyses of large areas and provide valuable

information on the behaviour of important components of the Wadden Sea area in the past in response to

changes that are expected to continue in the future, some of and which may have effects on coastal

defence structures, natural qualities and ecosystem resources and services

1 .3 Remot e Sens ingThe aims of this fore sighting study are to assess the possibilities and limitations of using remote sensing

as a tool for data monitoring relevant to the coastal defence objectives of the Delta Programme Wadden

Sea area. The intention is to generate knowledge on how the historical satellite data can provide valuable

information on developments of channels and gullies and sediment characteristics, in particular what

methods are available and applicable to the Dutch Wadden Sea. The focus was to use satellite data on: classification of sediments

identification and locations of mussel and oyster beds

detection of temporal changes in gullies and channels, which is considered as potentially relevant

for coastal defence issues in the Wadden Sea and for which remote sensing techniques and satellite

images are already developed and available.

Trend analyses will also reveal shortcomings in data sets and monitoring gaps that should subsequently

be addressed in the monitoring activities supporting the Delta Programme objectives. The advice on

relevant future applications of remote sensing in highly dynamic coastal seas such as Wadden Sea,

including new application of high resolution remote sensing, storm database and integrated model is

discussed.

1.3.1 Classification of sediments using satellite dataThe changes in sediment composition is considered to be strongly related to coastal erosion and climate

change (Bartholdy and Folving, 1986; Gomez et al, 1995; Hommersom et al, 2010; Srensen et al,

2006; Schmugge et al, 2002). Several studies using satellite Landsat data with a resolution of 30 meters

dealt with mapping the spatial distribution of sediment types in the German and Danish Wadden Sea.

However, methods were tested only in selected parts of the German and Danish Wadden Sea, which

differ considerably in parameters such as sediment composition and sediment transport (Postma, 1961;

Bartholdy and Folving, 1986; Srensen et al., 2006).

1.3.2 Ident ification and locations of m ussel and oyster bedsMussel and oyster beds, creating hard three-dimensional substrates and reefs (Fey et al., 2008), areconsidered important for coastal defence by reducing wave energy. Mudflats and mussel beds are

protected and recognised as useful for coastal defence along the Scottish part of the UK, as mentioned in

a National Marine Plan resolution of the Scottish Government, adopted in 16 March 2011(Scottish

Government, 2012). The National Marine Plan of Scotland is one of the main components of the Marine

(Scotland) Act, setting the national marine strategy, ensuring sustainable economic growth of maritime

industries, with taking the environment into account, and formulating policies on economic, social and

marine ecosystem objectives.

The classification algorithm on the identification of mussel and oyster beds using Landsat satellite data

developed by Brockmann Consult (Brockmann and Stelzer, 2008), is based on the habit of mussel and

oyster beds to strongly reflect sunlight because of a distinct colour and tone (dark) and a rough structure

of the surface, making these ecotopes very distinct from surrounding sediments.


9/45


1.3.3 Detection of t emporal changes in gullies and channelsThe temporal change in gullies and channels can provide information on coastal erosion processes

(Knighton, 1998; Gomez et al., 1995) and other developments which may affect the coastal integrity.Detection of temporal changes using satellite images is based on the analysis of series of images,

acquired by the same satellite, by the same sensor and in the same resolution. Only by doing so we will

be able to compare the details shown in the images.

1 .4 The research ob jec t i ves o f t h e s tudy1.4.1 The general objectiveThe general objective of this study was to assess the possibilities and limitations of Remote sensing

techniques as a tool for data monitoring relevant to the objectives of the Delta Programme Wadden Sea.

Additionally we wanted to explore the available expertise and knowledge on how historical satellite data

can provide information on developments of channels and gullies, morphology and sediment

characteristics.

1.4.2 The specific objectiveThe specific objective of this study is to answer the question what methods are available to provide

information on sediments composition, location of mussel and oyster beds and developments of channels

and gullies, which are considered relevant for coastal defence. The study specifically tests and identifies

methods which are applicable for the Dutch Wadden Sea.

To reach this specific objective, the following tasks were implemented: Review of relevant literature, satellite missions and available satellite data (resolution, optical

range), and properties of sediments, mussel and oyster beds, gullies and channels which can be

detected on satellite images. These issues are described in Section 3.2, Application of Remote

sensing in the Wadden Sea.

Applying the Brockmann Consult method on classification of sediments and mussel and oysterbeds, this is described in Sections, 3.3.1 Classification of sediments and 3.3.2 Classification of

mussel and oyster beds.

Detection of developments of channels and gullies, described in Section 3.3.3 Temporal changedetection in gullies and channels.

1 .5 The l im i t a t i ons o f t he s tudyAs the scope of this study did not allow analysis of the whole Dutch Wadden Sea, one location from which

sufficient ground data are available for a selection of the potentially relevant parameters was selected for

a pilot study, The parameters considered are: sediment characteristics, mussel and oyster beds and

temporal changes in gullies and channels. The analysis of the selected parameters started from selecting

the most applicable and relevant images from available satellite missions. For this reason an overview

was made of:

Satellite missions Periods of operation Revisiting frequencies (imaging dates and time) Used wave bands


10/45


The scope of this pilot study did not allow analysis and processing of all available (cloud-free and low

tide) images. Therefore it was decided to select the most representative images matching the dates of

classified images by Brockmann Consult, to compare the results and to see whether classification of

ecotopes in the Dutch Wadden Sea provided the same results, in the same way as had been done in the

German Wadden Sea.

1.5.1 Research methodologyThe focus of the Delta Programme determined the parameters selected for this study. The parameters

must be relevant for coastal defence and be detectable by satellite sensors. In general, there are several

satellite missions offering historical data from the Wadden Sea: Landsat (USA), ASTER (Japan), ERS (EU),

JERS (Japan) and SPOT (France). The Landsat missions have the largest continuous historical coverage

and resolution suitable to identify the selected parameters. One of the important preconditions for

developing useful Remote sensing procedures as a monitoring tool is the possibility to validate the results

of image interpretation with real ground data (ground thruthing). The current study employed a mixture

of different methodologies to classify sediments and mussel and oyster bed and to detect temporal

changes in gullies and channels from satellite images.

1.5.2 Structure of the reportChapter 3 on Materials and methods provides an overview on historical missions of medium resolution

satellites of Landsat, ASTER and ERS. The summary on missions available, type of sensors and data is

provided.

The sections on Classification of sediments, Mussel and oyster beds and Temporal developments in gullies

and channels deliver a general overview of Remote sensing classification studies implemented in the

Wadden Sea, the results obtained and experiences learned. The restriction of satellite data availability

(weather and tides, and usability of satellite missions) is presented. The inventory of satellite imagespresents images of different years and under different tidal conditions with the purpose to provide an

impression on how low and high tides images appear. The section on pre-processing of the satellite

images gives information on atmospheric correction and Principal Component analysis of images from

high and low tide. A test on Modified Soil Adjusted Vegetation index (MSAVI) is considered in relation to

the algorithm of Brockmann Consult. The usefulness of this index is explained, pointing out that it is

possible to detect mussel and oyster beds covered by macrophytes.

Chapter 4 describes the results of classifying sediment composition, mussel and oyster beds and temporal

changes in channels and gullies.

Chapter 5 on Future applications presents new developments on Remote sensing, including high

resolution WorldView- 2 image (2 meters in multi-spectral and 0.60 cm in panchromatic bands). The

experience from other marine institutions working with satellite data like the French Research Institute

for Exploration of the Sea (IFREMER) using historical radar ERS data, and a proposal for an Integrated

decision support model using satellite data, are presented.

Finally, the discussion outlines challenges and experiences gained from the study. For convenience of the

reader, Appendix 2 explains some technical terms used in Remote sensing science.


11/45


2 Ass ignmen tThe assignment of this study was to provide answers to:

Safety issues related to sediment budget and climate change of the Delta Programme WaddenSea, concerning the analysis of Development of channels and gullies and sediment

characteristics in the Wadden Sea based on an analysis of 30 years of satellite images and aerial

photographs available. Creating links to international literature and provide sediment data. The

research question of this component concerned literature review of relevant studies on properties

of sediments, mussel and oyster beds, gullies and channels which can be detected on satellite

images and selection and testing of algorithms.

What are the opportunities for use of Remote sensing?.


12/45


3 Mater ia ls and Meth odsRemote Sensing can be used to detect a wide variety of properties in the atmosphere and on the ground.

The development of image interpretation tools for specific parameters, which is a time consuming

process, is beyond the scope of this pilot study. The materials and methods discussed in this report apply

only to data analysed in relation to the objectives of the Delta Programme to assess the possibilities and

limitations of Remote sensing as a tool for monitoring the system based upon long-term historical

satellite data. The parameters selected to represent long-term changes in the Wadden Sea are sediment

composition, mussel and oyster beds and temporal changes in gullies and channels.

In Section 3.1 an overview of satellite data from Landsat, ASTER and ERS missions is presented. The

description on Period of operation, Resolution of the scene, Revisiting frequency, Purchasing details and

Wave bands (spectrum) of each mission is presented in Tables 1 and 2. The overview on Application of

Remote sensing in the Wadden Sea and the section on Classification describe methods used in Remotesensing studies implemented in German and Danish parts of the Wadden Sea. The Selection of pilot

parameters and pilot area where the classification method was tested highlighted the Factors restricting

satellite data availability. Pre-processing describes atmospheric correction of the images, which is

important in order to remove the effects of atmosphere in order to reduce the spectral reflectance of

sediments, mussel and oyster beds, and gullies and channels. To detect mussel and oyster beds covered

by macrophytes, the Modified Soil Adjusted Vegetation index (MSAVI) was tested.

3 .1 His to r i cal med ium reso lu t i on sa te l l i t e da taSatellites providing historical medium resolution images of the Wadden Sea area include missions from

the United States government agency National Aeronautics and Space Administration (NASA) and theEuropean Space Agency (ESA). The data from NASA Landsat satellites are going back to the 1970s,

providing long-term historical coverage of the entire globe (Figure 1- Landsat missions (NASA, 2011)).

The European Space Agency has cooperation with NASA and other space agencies, providing its multi-

mission ground systems to acquire, process, distribute and archive data from Landsat and other

satellites, known as Third Party Missions (TPM). The data from these missions are distributed under

specific agreements with the owners or operators of those missions, which can be either public or private

entities outside or within Europe.

Figure 1- Landsat missions (NASA, 2011).


13/45


3.1.1 LandsatThe Landsat missions provide extended large coverage of 170 km (north-south) by 185 km (east-west)

scenes on the ground. The revisiting time of satellite missions are every 18 (Landsat 1-3) to16 (Landsat5-7) days over the same area (705 km altitude, sun-synchronous, so that at any given latitude it crosses

directly overhead at approximately the same time each day). Landsat series were specifically designed to

detect characteristics about terrestrial features and wet and dry areas on land. The spectrum of different

bands ranges from visible blue (0.45-0.52 m), green 4 (0.52-0.60 m), red (0 0.63-0.69 m) and near

infrared (NIR; 0.77-0.90 m) to middle infrared (MIR; 1.55-1.75 m) bands. There are two additional

thermal infrared bands (10.40-12.50 m), measuring the radiant flux (heat) emitted from surfaces (land

and water) and mid-infrared, recording long wave data (2.08 - 2.35 m).

The Landsat missions employ among others MSS (Multi Spectral Scanner, Landsat 1-5), TM (Thematic

Mapper, Landsat 4, 5) and ETM+ (Enhanced Thematic Mapper Plus, Landsat 7) sensors. The images from

MSS have resolution from 80 to 120 meters in blue, green, red and near-infrared bands. The resolution of

TM is 30 meters in blue, green, red, near and mid-infrared bands with an additional thermal band in 120

meters. The ETM+ acquires data in all bands of the TM sensor and an additional high resolution 15

meters panchromatic band. In addition, the ETM+ sensor collects information in two different gain

settings (radiometry settings- image precision) in high and low mode. When brightness and reflection of

land surface is high, the sensor collects information in low gain, aiming to reduce size of collected data.

When surface brightness is lower, the sensor switches to high gain, in order to increase the capacity of

the sensor to collect as much details as possible. Table 1 provides an overview of Landsat satellite

missions. The main advantage of the Landsat (L4-7) mission is continuous temporal coverage over same

area for many years, using the same sensor and spectral range in resolution of 30 meters, with a latest

addition of 15 meters panchromatic band. The recent failure of Landsat-7 data mission closed provision of

data in October 2011, however, a new mission of the Landsat-8 is planned in the next two years, which

will continue global acquisition. The detailed technical characteristics of the Landsat data on acquisitiondates and position of the sun, cloud condition and sensor parameters can be found in metadata files

which are available for each image.

3.1.2 ASTERASTER is a cooperative effort between NASA and Japan's Ministry of Economy Trade and Industry (METI),

in collaboration with scientific and industrial organisations in both countries. ASTER captures high spatial

resolution data in 14 bands from visible to thermal infrared wavelengths and provides stereo viewing

capability for digital elevation model creation. The ASTER satellite was launched in 1999 and can provide

information in 15 meters resolution and as the Landsat satellite has global coverage with temporal

acquisition every 16 days of the same area covering 60 km on the ground. The advantage of ASTER

images is its high resolution, making it possible to enhance Landsat data, because of the same spectraland temporal resolution (Table 2).


14/45


Table 1- Overview of the Landsat satellite missions

Sate l l i te Landsat 1-3

(RVB, MSS)

Landsat 4 , 5

( TM, MSS)

Landsat 7

ETM+Per iod of

opera t i on

1972 - 1983 TM: 1982

MSS: 1982 - 1995

1999 -

Resolut ion

scene

57 x 79 m

170 x 185 km

30 m (120 m)

170 x 185 km

30 m (15 m, 60 m)

170 x 185 km

Rev is i t ing

f requency

18 days 16 days 16 days

Purchas ing

de ta i l s

ESA: until 2010 free of charge for the scientific community.

Wave bands

( s p e c t r u m )

RGB (Red, Green, Blue)

sensor

1:475-575 nm- B

2:580-680 nm- G

3:690-830 nm- R

Multi Spectral Scanner

(MSS) sensor

4:0.5-0.6 m

5:0.6-0.7 m

6:0.7-0.8 m

7: 0.9-1.1 m

8:10.4 12.6 m (L3)

Multi Spectral Scanner

(MSS) sensor

4:0.5-0.6 m

5:0.6-0.7 m

6:0.7-0.8 m

7:0.9-1.1 m

Thematic Mapper

(TM) sensor

1:0.45-0.52 m2:0.52-0.60 m

3:0.63-0.69 m

4:0.76-0.90 m

5:1.55-1.75 m

6:10.40-12.50 m (120m)

7:2.08-2.35 m

Enhanced Thematic

Mapper Plus

(ETM+) sensor

1: 0.45-0.52 m

2: 0.52-0.60 m

3: 0.63-0.69 m

4: 0.76-0.90 m

5: 1.55-1.75 m

6: 10.40-12.50 m (60m)

7: 2.08-2.35 m

8: 0.52-0.90 m (15 m)

3.1.3 ERSThe ERS-1 satellite is one of the oldest European satellites, with global coverage dating back to 1991. T

Compared with the Landsat and ASTER missions, the ERS-1 satellite provided data in spatial resolution of

12,5 meters with coverage of 5 km by 5 km, with three planned repeat cycles of 35 days. The ERS-1

provided data in black and white mode, and compared with the Landsat and ASTER satellites, the data is

not influenced by clouds and weather, because of the ability of radar system (pulse signal sending and

receiving) to penetrate clouds and rain. The ERS-1 was joined in 1995 by ERS-2, providing tandem

missions until 2000, finally stopping the ERS-2 satellite because of satellite control failure in July 2011.

The archived data of the ERS-1 and 2 satellites are available from ESA, free of charge for the scientific

community. An overview of ERS-1 and 2 satellite missions is given in Table 2.


15/45


Table 2- Overview of ASTER and ERS satellite m issions

Satel l i t e ASTER ( NASA + Japan RS) ERS-1 and ERS-2 ( EU)

Per iod o foperat i on

1999 to present 1991 2011

Resolu t i on (and

coverage)

15 m (60 x 60 km) 12.5 m (5 x 5 km)

Revis i t ing

f requency

16 days Revisit depends from the area,

starting from 3, 35 and up to 168

days. European area revisit time is

from 3 to 35 days

W av e bands

( s pec t r um )

Blue 1 (0.45 -0.52 m)

Red 3 (0.63-0.69 m)

Near infrared 4 (0.76 - 0.90 m)

Short-wave infrared 5 (1.55-1.73 m)

Thermal infrared 6 (10.4-12.5 m)

Radar panchromatic images

3 .2 App l i cat ion o f Rem ote sens ing in th e Wadden SeaBrockmann Consult analysed 20 years of historical data of Landsat TM (60 and 30 meters resolution)

starting from 1987, focusing on the German Wadden Sea and the province of Zeeland in the Netherlands.

The surface mapping used data from high resolution ASTER images and measurements from an air plane

(HyMap imager). The results were validated using in-situ measurements. The following characteristics of

tidal flats were detected from satellite images with reasonable accuracy:

sediments composition (sand and mud)

changes in locations of vegetated areas (macrophytes) using multi-temporal images

location of mussel and oyster beds.

The Danish part of the Wadden Sea was mapped in 2006 using Landsat ETM+ (30m) data. For sediment

classification, combinations of several bands were analysed, and best combinations were found in blue

(0.45 -0.52 m), red (0.63-0.69 m), near infrared (0.76 - 0.90 m), short-wave infrared (1.55-1.73

m) and thermal infrared (10.4-12.5 m) (Srensen, et al, 2006). The final results showed a fairly good

distinction of sediments in 4 main classes: mudflat, mixed flat (wet/moist), low (wet) sand flats and high

(dry) sand flat.

Geospatial Data Service Centre (GDSC, Haartsen and Marrewijk, 2001) analysed temporal changes in

movements of sandbanks and tidal channels in the Wadden Sea using Landsat MSS images recorded

between 1975 and 1987. The satellite data were combined with in-situ measurements and an attemptwas made to detect reflectance properties of sediments and biogenic structures (Curran and Novo, 1988;

Schmugge et al., 2002).

NIOO-CEME implemented several studies on monitoring sediments grain size in the Westerschelde

(southwest Netherlands), based on information from radar and optical Remote sensing (Van der Wal and

Herman, 2007; Van der Wal et al., 2005; 2008). Three algorithms were combined exploring:

a) properties of radar images

b) reflectance properties of sediments in optical (green) and short infra-red bands

c) combination of these two methods with in-situ measurements.

Van der Wal et al. (2008) classified the distribution of intertidal macro benthos using Remote sensing and

in-situ measurements. The aim of the study was to assess distribution of macrobenthos and sediments in


16/45


relation to dynamic processes in the Westerschelde. The study stated necessity of validation and

calibration in order to use this method in other areas.

Overall, Landsat images with 30 meters resolution (Landsat 4-7) were better in classification of

sediments and location of biogenic structures, compared with images of lower MSS (60 meters) resolution

(Landsat 1-3).

3 .3 Classi f icat ion3.3.1 Classification of sedimentsBartholdy and Folving (1986), Haartsen and Marrewijk (2001) and Srensen et al. (2006) described

methods on classification of sediments using Landsat TM and ETM+. The methods are based on

peculiarities of sediments with different composition of sand, clay, silt and water content to reflect distinct

spectral signatures which can be recorded by satellites. The data processing in this case will do spectralunmixing for detecting and separating spectral signatures corresponding to each sediment type.

However, the method of spectral unmixing in the Wadden Sea is of limited use because of the high

nonlinearity of sediment transport and gradients in sedimentation processes, patterns of high temporal

and spatial variability, and tidal asymmetries (Stanev et al., 2007; Hommersom et al., 2010). The

processes in the Wadden Sea produce non-correlated patterns of sediment mixtures, and it is different

not only from one image to another, but even within one image. The spectral unmixing in this case is not

always able to differentiate very precisely different objects and substances, potentially resulting in

reduced classification accuracy and mismatches. Validation of final results with field data, ideally acquired

during or within a very short time lag after satellite data acquisition, will be extremely important to refine

classification. Classifying Wadden Sea sediments by grain size is limited in medium resolution satellite

data (Bartholdy and Folving, 1986). The final classification created six main groups of water, mudflat,

muddy or mixed flat, wet or moist sandflat, dry sandflat and high-sand, without going into further details.

Another limitation is temporal change detection using images from different dates. The satellite sensor

records momentum states of tidal flats and each image has its own specific non-correlated pattern of

temporal and high spatial variability, further complicated by the periodicity of the tidal wave. This means

that reusing spectral signatures from one image to another is not always possible.

Brockmann Consult tested a spectral unmixing technique in German and Dutch parts of the Wadden Sea

and in Zeeland. In addition, the method is enhanced by adding high resolution ASTER data and supplying

textural information from ERS and TerraSAR-X (Germany) radar data. The advantage of the developed

method is the possibility to do quantitative analysis of satellite images without prior knowledge of thearea and to predict sediment composition (content of mud and sand) and location and size of areas

covered with macrophytes, using spectral libraries from reference images.

3.3.2 Classification of mussel and oyster bedsBartholdy and Folving (1986), Brockmann and Stelzer (2008), Fey et al. (2008) and Hommersom et al.

(2010) described a classification method for mussel and oyster beds. The detection of intertidal mussel

beds is based on spatial characteristics of mussel and oyster beds in tone and colour (dark appearance

compared to bare sediment), shape (stretched patterns along the tidal flats), texture (rough surface),

size (irregular rounded) and shadow (originated from sediment deposits from mussels and oysters). The

Normalized Vegetation Index NDVI detects mussel and oyster beds covered by macrophytes. The method

applied by Brockmann Consult used the same prediction algorithm described for sediment classification.


17/45


Textural information from radar images served as added information in delineation of mussel and oyster

beds when beds are slightly covered by water or sediment.

3.3.3 Temporal change detection in gullies and channelsThe location of gullies and channels and their movements in course of time can be detected using 30

meters resolution Landsat TM images. As revealed in studies by Postma (1961), Bartholdy and Folving

(1986), Srensen et al. (2006) and Hommersom et al. (2010), the composition of sediments and their

transport and spatial distribution are different in every location of the Wadden Sea. In order to detect

changes it is important to use satellite data taken in the same optical range. The algorithm for change

detection computes differences between two images. The spectral values from the first image are

subtracted from the second image. The next step is to compute statistical variability for both images and

highlight differences (change). This algorithm used Landsat images from different dates.

3 .4 Se lect ion o f Remote sens ing c lass i f i ca t ion m ethodBrockmann Consult, with support from the European Space Agency developed a prediction algorithm to

detect sediment characteristics, changes in vegetated areas (macrophytes) on emerging tidal flats and

locations of mussel and oyster beds from Landsat images using field information on water content and

water coverage, location of mussel and oyster beds and sediment composition. The first step of the

algorithm was to separate intertidal flats from water. The next step was to separate more details of

intertidal flats. The core of the Brockmann algorithm followed a decision tree rule, with specific

differentiation (Figure 2). To detect mussel and oyster beds covered by macrophytes, the method of

vegetation indexing is tested.

Figure 2- Brockmann Consult designed decision- tree rules


18/45


In cooperation with Brockmann Consult it was decided to use its available expertise and to apply its

method to see whether it is also applicable for the Dutch part of the Wadden Sea

3 .5 Se lect ion o f p i l o t pa ram ete rs and p i l o t a reaSediment properties and sedimentation processes are important parameters in relation to coastal erosion

and coastal defence issues in the Wadden Sea (Madsen et al., 2010). For the Delta Programme Wadden

Sea, aspects affecting sediment dynamics are of particular interest. The selection of parameters for this

pilot study considered features that are presumed relevant in terms of sediment budgets in the Wadden

Sea. In addition, the parameters should be detectable by satellite sensor and field data should be

availability for validation.

The changes in sediment composition, location of mussel and oyster beds and gully and channel

development were considered as relevant for the Delta Programme Wadden Sea. Mussel beds can catchlarge amounts of sediment influencing its own and nearby surroundings. By doing so they act as

sediment accumulating and storing systems and at the same time play a role in reducing wave energy

during storms. The stored sediment may later become available for salt marshes (Oost, 1995). There are

several satellite missions providing data acquisitions on global and regional scales, including the Wadden

Sea. The medium resolution (30 meters) Landsat 5-7 missions over Wadden Sea started from 1984 and

have continues acquisition till 2011. This allows comparison of different images taken in the same optical

range and by the same sensors and sufficient time span between images to analyse temporal

developments. The other satellite missions data can serve as important supplementary information to

verify and to improve the final classification results.

The Schiermonnikoog - Rottumeroog Islands area was selected as pilot location (Figure 3). The

advantages of this area are:

Relatively undisturbed environment (area closed for fishing since 1990s) Limited area: limited tidal variation Availability of temporal and spatial coverage of ground data for the selected parameters for validation Links to other DP-Wadden projects (sediment dynamics and system knowledge) Potential interest from e.g. other DP-Wadden project partners and Brockmann Consult.

It is assumed that the selected area is representative for the Wadden Sea and that the results and

developed Remote sensing techniques may be extrapolated from this area to cover the whole Dutch part

of the Wadden Sea, and beyond.


19/45


Figure 3-Location of pilot area (Rottumeroog and Schiermonnikoog islands) (Right- Netherlands map; satelliteimage of Schiermonnikoog and Rottumeroog I slands).

3 .6 Fac to rs res t r i c t i ng sa te l l i t e da ta ava i l ab i l i t y - w ea the r and t i desThe main requirements for using optical, multi-spectral satellite data for (intertidal) developments in the

Wadden Sea are cloud free and low tide conditions. Climate data from station Lauwersoog showed that on

average, there are 10 days per month with more than 50% maximum potential sunshine duration (data:

KNMI).

The tidal differences measured at locations nearest to Schiermonnikoog and Rottumeroog islands at

Huibertgat, Netherlands (53.5667 N, 6.4000 E), Lauwersoog (53.4167 N, 6.2000 E), Eemshaven,

Netherlands (53.4500 N, 6.8333 E) and Borkum, Germany (53.5833 N, 6.6667 E) presented tide

information for every image used in the study. The main challenge caused by tidal movements is the

difference in harmonics of tidal waves within one image, affecting classification accuracy. Tidal

movements over four stations in the investigated area are presented in Figure 4.

Although possibly these tidal predictions may be different from the actual sea level because of additionalfactors such as wind, atmospheric pressure etc, the information is still helpful to select potentially usable

satellite images.


20/45


Figure 4 - Tidal movem ents over the stations of Huibertgat (H), Schiermonnikoog (S), Eemshaven (E).

3 .7 I n v e n t o r y o f s a t e l li t e i m a g e sLandsat revisits the Wadden Sea area every 16th day, in morning between 09.30am and 10.30 am GMT.

Compared to the low water table, this means that all passages occur during daytime but less than 50% of

the passages occur at low tide. Combining the satellite revisiting scheme and tidal charts with weather

data, on average 4 low tide and cloud-free images are available from each satellite mission every year for

the Wadden Sea. Landsat uses 233 paths (orbits) for total global coverage. Path numbers 197, 198 and199 cover the eastern, middle and western parts of the Dutch Wadden Sea, respectively. A complete

overview of available cloud-free and low-tide satellite images is given in Appendix 1.

3 .8 Pre-process ing3.8.1 Atmospheric correction of Landsat im agesThe atmospheric correction of satellite images is necessary as it reduces the effects from moisture

evaporation, haze and other disturbances generated by the atmosphere. The atmospheric correction

removes effects from atmosphere using time and date, sun elevation, azimuth of sun illumination and

azimuth angle during image acquisition, to make corrections for each wave band. The disturbance from

satellite motion is reduced by applying corrections called gain offset.

The formula on atmospheric correction for Landsat images is (Chander and Markham, 2003):

Where: L- spectral radiance (in Watt)

Qcal- quantized calibrated pixel value in Digital Units (DN)

Qcalmin- minimum quantized calibrated pixel value in Digital Units (DN), which is equal to 0 (zero).

Qcalmax- maximum quantized calibrated pixel value in Digital Units (DN), which is equal to 255 (two

hundred fifty five).

LMAX- spectral radiance that is scaled to Qcalmax

LMIN- spectral radiance that is scaled to Qcalmin.

Area starting

high tide

Area with

low tide

E

H

L

S


21/45


The atmospheric correction visibly improves contrast of the images (Figure 5).

Figure 5-Landsat-5 (TM) image of Rottum island, date of acquisition- 1995-11-25, before (left) and after (right)

atmospheric correction.

3.8.2 Spectral analysisThe spectral separability is analysed using the technique of Principal Component Analysis (PCA). In

addition, Normalized Vegetation Index (NDVI) was derived to see if any biomass (vegetation) is covering

emerged tidal flats. Figure 5 present results of PCA analysis. Differences in sediment concentrations and

turbidity of the water can be seen around Rottumeroog island and in the Eems-Dollard estuary expressed

in different colours. To see the differences in images from tide level, images in low and high tide were

used. Images in low tide from 1995-11-25, 1990-03-05; 2000-05-13; 2006-09-12 and almost low tide

from 2009-07-01; 2010-07-04 show detailed composition of tidal flats, expressed in variations of

different colours. Two images were selected for classification, one from 1990-03-05 because it is showing

large areas of dry tidal flats and one from 2006-09-12, because it is matching with classified images in

the German Wadden Sea from Brockmann Consult. The images from 1995-11-25, 2009-07-01 and 2010-

07-04 were used to detect changes in a long period of 15 years (from 1995 to 2010) and in short one

year period (from 2009 to 2010).

Table 3 shows the scenes that were selected for this pilot study, images in low tide, with two images used

in classification of sediments and mussel beds and Table 4 image in high tide, not used in classification.

In order to analyse the effect from tidal level the spectral analysis (PCA) is presented both for images in

low and high tide.

The images from high tide in Table 4 are given for general overview and PCA images are explains whyimages in high tide cannot be used for classification. However, images in high tide still can be used to

detect changes along the main coast and coast of the islands.


22/45


Table 3-I mages in low tide selected for t his report

Landsat-5 TM Low tide

Image acquisition date: 1995-11-25 Tide information by station Lauwersoog (53.4167 N, 6.2000 E):

-Time of High tide- 12:02 PM CET/ 2.91 meters

Original im age

In original image mainland and islands is

shown in red color, because of reflectance of

chlorophyll of vegetation in red and infra-red

bands. The water from the North sea in deep

blue color. The dry area of tidal flats in the

Wadden Sea presented in brown and light

blue color.

PCA image

The PCA analysis shows that there are

disctict differences between major

structures, like sediments (green), sea

water (violet) and sandy beaches (very

bright yellow).

Landsat-5 TM - Low tide


- Time of Low Tide -9:50 AM CET /0.66 meters

- Time of High Tide- 4:06 PM CET /2.49 meters

Original im age

One of the images selected for sediments

classification. Sediments in a high contract

with coastal land and from gullies andchannels.It is one of the very few best

images showing such high separability of

sediments from coastal land and gullies.


23/45




-Time of High tide-6:59 AM CEST / 2.29 m

-Time of Low tide- 12:39 AM CEST / 0.26 m

Original im age

The image covers large area of the Wadden

Sea and low tide can be observed in part of

Schiermonnikoog and Rottumeroog islands.

The Texel island area showing situation of

high tide. The image presents short time lag

in tidal variations, which can be seen within

the islands.

PCA image



structures, like sediments (green), sea

water (ligher green) and sandy beaches

(very bright yellow).

Landsat-7 ETM+- Low tide


-Time of Low tide- 8:24 AM CEST/ 0.19 meters

Original im age

One of the images selected for mussel beds

classification. The date of image acquisition

matches with classified image in the German

Wadden Sea by Brockmann Consult.


24/45



Image acquisition date: 2009-07-01 Tide information by station Lauwersoog (53.4167 N, 6.2000 E):- Time of High tide- 5:50 AM CEST / 2.32 m- Time of Low tide- 11:44 AM CEST / 0.39 m

Low tide in the eastern part of the Dutch Wadden Sea, incoming tide in the western part

Original image

Image presents very high tide level in Texel

island, with low tide in Schiermonnikoog and

Rottumeroog islands. The mainland coast

and islands are clearly delineated, with

channels and gullies around

Schiermonnikoog and Rottumeroog islands.

PCA image



structures, like sediments (pink), sea water

(greenish) and sandy beaches (very bright

almost white). The cloud on top of the

image is very light green).

Landsat TM-5- Low tide

Image acquisition date- 2010-07-04 Tide information by station Lauwersoog (53.4167 N, 6.2000 E):- Time of High tide- 4:00 PM CEST / 2.50 m- Time of Low tide- 9:47 AM CEST / 0.43 m

Low tide in the extreme eastern part of the Dutch Wadden Sea, more or less high tide in the western

part

Original im age

Image a bit distorted by haze from clouds,

spread over islands. The distinction of

mainland coast and islands are very clear

and water in the Wadden Sea showing

shallow areas of tidal flats.

PCA image



structures, like sediments (pink), sea water

(greenish) and sandy beaches (very bright

almost white). The cloud on top of the

image is very light green).


25/45


ASTER image- High tide

Image acquisition date: 2007-06-08 Tide information by station Lauwersoog (53.4167 N, 6.2000 E):- Time of High tide- 8:34 AM CEST / 2.34 m- Time of Low tide- 2:39 PM CEST / 0.52 m

Original image

The ASTER image is high tide, showing very

clear the coast of mainland and islands. The

tidal flats are completely covered by water

and it is difficult to see channels and gullies

development.

The PCA on ASTER image not done, because

ASTER image in high tide and therefore it was

not considered to use.

Table 4- I mages in high tide, not used in the study

Landsat- 5 image- High tide

Date of acquisition- 1989-05-23 Tide information by station Lauwersoog (53.4167 N, 6.2000 E):- Time of High tide- 11:25 AM CET / 2.43 m

Original im age

Image in high tide showing high level of water

around Schiermonnikoog and Rottumeroog islands.

The delineation of mainland coast and islands are

very clear. There are differences in color of the

Wadden Sea and North Sea, which can be

explained by bathymetry (deep in the North Sea)

and shallow and turbid waters in the Wadden Sea.

PCA image

The PCA analysis shows that there are disctict

differences between major structures, like

sediments (green), sea water (violet) and sandy

beaches (very bright yellow). However, tidal flats is

completely covered by water and sediments

expressed in green color are detected because

area of the Wadden Sea is shallow and turbid.


26/45


Landsat-5 image- High tide

Image acquisition date- 1996-01-19 Tide information by station Lauwersoog (53.4167 N, 6.2000 E):- Time of High tide- 8:50 AM CET / 2.63 m

Original im age

The channel and gullies development are not very

well detectable. However, sandy beaches on

islands and salt marshes along the mainland coast

are clearly delineated.

PCA image

The PCA analysis shows that there are disctict

differences between major structures, like

sediments (green), sea water (violet) and sandy

beaches (very bright yellow). The Wadden Sea is

completely covered by high water and there is no

much details are present.

The results from PCA analysis in one selected image show direct correlation of sediment reflectance and

water content (caused by tidal differences). The scatterplot of the PCA image (Figure 6) shows highest

amount of data in blue, (Band 1; 0.45 - 0.52 m), green (Band 2; 0.52 - 0.60 m) and red (Band 3;

0.69-0.83 m), which is characteristic for water, wet soil and vegetation respectively.

Figure 6- Scatter plot of PCA image (left) and corresponding PCA image (right)


27/45


Table 5 presents PCA statistic, with very high eigenvalues (PCA values) in first 3 bands and minimum

values in band 6, which is mid-infrared band.

Table 5- Statist ics of PCA analysis re Figure 6

Bands Min

(lower red line)

Max

(upper red line)

Mean

(grey line)

Stand ard dev ia t i on

(green lines)

Eigenvalue

Band1 163 188 179.147361 4.921281 435.492695

Band2 172 197 184.440027 5.495821 83.351276

Band3 97 106 102.328307 2.674595 34.799761

Band4 52 87 69.113777 6.580676 10.199938

Band5 26 89 55.136395 14.800777 4.254306

Band6 19 85 45.261823 15.628093 0.084134

3.8.3 Result from Modified Soil Adjusted Vegetation indexThe other spectral analysis on Modified Soil Adjusted Vegetation index (MSAVI), which is correlation of

near-infrared NIR and red bands containing soil brightness correction factor. The difference between

MSAVI and traditional vegetation and soil indexes, that MSAVI contain both indexes.

Equation 1 - Soil bright ness correction factor

Where: NIR- near infrared, RED- red bands; s- slope of soil line computed from plot of red versus near-infrared brightness values.

Equation 2- Modified Soil Adjusted Vegetation index

Where: NIR- near infrared, RED- red bands; L- soil brightness correction factor.

The index has values from minimum 0 (no contrast for soil, usually areas with water) to maximum 1

(depending from soil composition- mud, silt and sand content). Visually variations can be seen from very

bright high contrast, usually it is sandy and or very dry soil and very dark- low contrast- normallyrepresenting soil with high mud or water content. Results of MSAVI index in Figure 7. The image shows

low contrast in soil values in emerging tidal flats (very dark grey) compare with high contrast sandy coast

of the islands (very bright).


28/45


Figure 7-Landsat - 7 ETM+ ; acquisition date- 2000- 05-13

3.8.4 Radar imagesThe radar ERS data provides information on texture and roughness of surface features. The properties of

ERS series satellites can provide additional information, for instance to verify classification results from

Landsat data. One of the possible applications of radar data is image fusion. Images in lower resolution,e.g. Landsat 30 meters, can be merged at pixel level with ERS radar 12.5 meters data. As a result, a

Landsat image will keep its multi-spectral properties which are useful in classification and will have an

improved spatial resolution from 30 meters to about 20 meters. Figure 8 presents ERS-1 data of the

Schiermonnikoog and Rottumeroog islands, compared to Landsat.

Figure 8-ERS-1 radar data of Schiermonnikoog and Rottumeroog islands. Radar image (left) with delineated

area of islands. Landsat image on right, showing the same area.

The radar image shows land in dark grey and water in light grey. Although the image is taken during low

tide on 2003-10-28, the intertidal flats are not clear in the image.


29/45


4 Resul ts4 .1 Classi f icat ion4.1.1 SedimentThe results of sediment classification using Landsat TM (30 meters resolution) data combined with ASTER

and ERS (Figure 9) showed a clear separation between sand classes, mixed sediments, silt and silt mixed

sediments and some mussel and oyster beds. The algorithm has been developed to predict sediment

concentrations (content of mud and sand), location and size (area) of mudflats and areas covered with

vegetation. The overall accuracy of the method comparing classified sediments with field data was nearly

80%.

Figure 9-Sediments classification. Sediment types (in various colours) in the area between Rottumerplaat-

Rottum eroog and the Frisian coast. Yellow = sand, purple = silt, gr een= vegetation, blue = water. I mage date

from 1990- 05-0 3. Circles indicate locations for ground t ruthing. The red circle indicates a mismatch between t heclassification result and field data.


30/45


The table in Figure 9 shows the area size for each classified sediment type, expressed in hectares.

The red circle in Figure 8 indicates a mismatch with field data. Very muddy areas with high water content

can be mismatched with sandy areas also containing water, as they may appear similar. The dynamic of

tidal movements and the water content of the sediment may also play role, since the area in the red

circle (sandy according to field data) was already starting to fill up with water, whereas the areas

indicated with the green circles were still dry. This means that satellite images have a high accuracy

when there is a perfect match between low tide and the time the image was made. However, if there is a

time difference the accuracy decreases and PCA images presented in Table 3 and 4 showing the distinct

details in images in low tide and not much details in images in high tide.

The sediment image (Figure 8) dates from early May from 1990. Mussel beds were not present in that

year. During early spring similarity in green colours in vegetated areas in the saltmarshes and some parts

of sand and mudflats and arable land does not necessarily imply a similarity in chlorophyll, and/or

vegetation biomass or coverage.

4.1.2 Classification of mussel/oyster bedsFigure 10 shows the result of a satellite image processed using the Brockmann algorithm developed to

classify mussel and oyster beds. Superimposed over this image are the locations of mussel and oyster

beds based on IMARES field data. The classified image from September 2006 shows mussel beds in a

dark green colour. The black surrounded polygons in this figure represent a compilation of IMARES field

data on the occurrence of mussel beds in 2005-2007, when large areas were covered with mussels.

Mussel beds developed in this are from 1994 onwards after their disappearance in 1990 due to fisheries.

A major expansion occurred in 2001 and after 2007 a gradual decline occurred. The light green colour

may indicate fine silt which may cover parts of the bed area.

Class_ Names Color

Unclassified

Sand classes

Sand classes 1

Sand classes 2

Sand classes 3

Silt sediments1

Silt sediments 2

Probable mussel beds

Mussel beds

Other water features

Figure 10- The location of mussel and oysters beds in the Simonszand-Frisian coast area (presented in darkgreen). Yellow = sand, purple = silt blue = water. Black lines indicate the m aximum extension of mussel and

oyster beds present between 2005 and 2007, based on IMARES field data. The satellite image represents the

situation in 2006.


31/45


The Remote sensing classification result from 2006 shows a good match with the field data from 2005-

2007. Some possible mismatches may be explained by tidal dynamics in one image, which may affect the

appearances (spectral signatures) of the mussel and oyster beds. IMARES field data suggest that some

mussel and oyster beds are located below the water line at the moment the image was recorded. In

general the results are very promising.

4.1.3 Temporal development of gu llies and channelsThe intention of our study was also to generate knowledge on how historical satellite data can provide

information on developments of channels and gullies and sediment characteristics, in particular what

methods are available and what the applicability of such methods are to the Dutch Wadden Sea.

Gully and channel developments have been analysed by comparing two images using a change algorithm

which can be used for historical monitoring of optical Landsat TM (30 m) images. The channel dynamics

can be visualised and interpreted.

The main limitation of trend analysis specifically for the Wadden Sea is the high dynamics of tidal flats,

recorded on satellite images as momentum state of non-correlated temporal patterns in high spatial

variability, further complicated by the periodicity of tidal waves. The study by Hommersom et al. (2010)

revealed that existing algorithms tested in the Wadden Sea are limited due to a high variation in

concentrations of various substances and sediments. It is difficult and not always possible to precisely

differentiate spectral mixes of different objects and substances present in one area. In this case a

comparison of images for specific trend analysis of specific patterns in sediment composition is not

possible. However, it is possible to see such trends in hard substances such as coastal land, which is

clearly demonstrated in Figure 11, showing changes in the course of channels as well as changes in the

shape of the sandy coastline on the northern shore of the island of Schiermonnikoog, Rottumeroog and

Rottemerplaat between November 25, 1995 and July 1, 2009.

Especially the inner delta between Schiermonnikoog and Rottumerplaat shows strong changes in the

running of channels as well as a widening of some channels. Early detection of such changes may be very

helpful in detecting potential threats for coastal defence. Changes over one year are presented in Figure

12, comparing images of July 1, 2009 and July 4, 2010. The two arrows indicate erosion at Simonszand

(east) and De Balg at Schiermonnikoog (west).


32/45


Figure 11-Upper image: changes in morphology between November 25, 1995 (lower image left) and July 1,

2009 (lower image right). The level of change is indicated by the amount of contrast. Grey areas indicate nochange. Darker areas indicate features present in the past (i.e. 1995) and could point to erosion; lighter areas

indicate features that were not visible in the past and could point t o newly emerged int ertidal areas.

Figure 12-Upper image: changes in morphology between July 1, 20 09 ( lower image left) and July 4, 2010 (lowerimage right). The level of change is indicated by the amount of contrast. Grey areas indicate no change. Darker

areas indicate features present in the past ( i.e. 2009) and could point to erosion; lighter areas indicate featuresthat w ere not visible in the past and could point to newly em erged intertidal areas.

Changes be tw een November 2 5 , 1995 and Ju l y 1 , 2009

November 25 , 1995 Ju ly 1 , 2009

Changes be tw een Ju l y 1 , 200 9 and Ju l y 4 , 2010

Ju ly 1 , 2009 Ju ly 4 , 2010


33/45


5 Fut u r e app l i ca t ions5 .1 Sto rm da tabaseAn application with relevance for decision-making with respect to coastal defence is a storm database.

IFREMER (the French Research Institute for Exploration of the Sea) developed a storm database based on

historical Remote sensing data (ERS-1 and ERS-2) and information on storm history:

date, time duration of the storm in hours strength of the storm: wind direction; wave height (maximum in meters); wind velocity (meters

per hour); amount of rain (in mm)

location and size of affected area, derived from historical ERS data.The meteorological data are used for deriving a Storm occurrence index, estimated by statistical analysis

of historical storms and storm forecasts. The Storm occurrence index shows the probability of future

storms, and models areas to be affected. A storm occurrence index is very useful in combination with

information on stability and resilience of coastal ecosystems which have a wave attenuating capacity such

as mussel and oyster beds and salt marshes. Stability and resilience of these ecosystems in relation to

storm occurrences (strengths, frequencies, direction of attack) determine the extent to which coastal

defence strategies can rely on wave attenuation by salt marshes and mussel and oyster beds. A storm

damage observation team (a saltmarsh resilience team or Quick Reaction Force) is recommended to

gather field data on storm damage and subsequent recovery time for wave attenuating coastal

ecosystems. A Quick Reaction Force has the objective of gathering relevant information and data on

coastal damage as soon as possible after a (major) storm. Base-line data on coastline, dunes, locations

and extents of mussel and oyster beds, salt marshes and channels and gullies, and elevation of intertidal

flats and salt marshes need to be available before each storm event. Re-establishment needs to bemonitored between storm events and calibration before the next storm event is necessary to reset the

base-line data. Organisation of pre- and post-storm measurements will determine the pace at which

relevant (eco) systems recover from storms and if the measures taken as to compensate for erosion are

sufficient to meet the objectives of the Delta Programme and to uphold the pre-set safety levels.

Measurements should comprise new dune and tidal marsh formation between two storm seasons as to

determine the (semi) natural resilience of dune and tidal marsh systems to storm events. Remote sensing

may prove instrumental to assess these data just before and immediately after each storm season.

5 .2 I n t e g r a t e d m o d e lAnalysis of the past provides scenarios for future developments on which coastal defence strategies canbe based. Remote sensing data and meteorological forecasting can be merged into an integrated model

with the objective to predict possible extreme event (storm) damage to e.g. intertidal flats, mussel and

oyster beds, salt marshes, dunes and beaches, etc. For the development of such a model the expertise

and instruments of e.g. IFREMER, Deltares and IMARES can be combined (Strom index, FEWS- Water

Level Forecast Early Warning System, IMARES Wave prediction mode). This integrated model can be

updated each year with new information on the developments of gullies and channels, intertidal flats, salt

marshes and mussel and oyster beds, fine-tuning the short term predictions and extending long term

projections. Such a decision support tool will be a powerful instrument for coastal managers as to

evaluate the effectiveness and longer term sustainability of their management practises and if and when

adaptations are necessary to meet certain coastal defence policy standards. The proposal on an

integrated model is presented in Figure 13.


34/45


Figure 13-Proposal of integrated decision support model

5 .3 High reso lu t i on sa te l l i t e da taThe new generation of high resolution satellite missions include sensors with imaging capability in

resolutions better than 10 meters. The WorldView-2 images cover the Frisian community of Dongeradeeland the north-western part of Groningen, De Marne and partly of the islands Schiermonnikoog,

Rottumerplaat and Rottumeroog islands (Figure 14).

Figure 14- WorldView-2 images

High resolution WorldView-2 images contain information in blue, green, red and infra-red bands with an

additional panchromatic very high resolution (0.60 cm) band. The data is supplied by the commercial

company Digital Globe (USA). The standard image product contains radiometrically corrected pixels.


35/45


Further corrections for atmospheric and other effects are necessary. An overview of sensor characteristics

is given in Table 6.

Table 6- Sensor characteristics of the WorldView-2 satellite product information

Product

Multispectral (4 -bands) Resolution-2,4 meters

Bands- Blue, Green, Red, NIR1

Panchromatics band Resolution- 60 cm

Spect ra l reso lu t ion

Multispectral (4- bands) Optical range

Blue 442 - 515 nm

Green 506 - 586 nm

Red 584 - 632 nm

NIR1 765 - 901 nm

Panchromatics band 447 - 808 nm

Compared with medium resolution data (e.g. Landsat), high resolution WorldView-2 data has shorter

extents in blue, green and red spectrums, meaning that collected information is less comparable with the

Landsat sensor, but this is compensated by the advantage of high resolution. The 0.60 cm panchromatic

band is ideal for visual interpretation, especially to identify and to locate mussel and oyster beds.

An automated detection algorithm for location and size of mussel and oyster beds and sediment

composition is being developed by IMARES (submitted for publication). The algorithm is based on building

libraries of spectral signatures of sediments and mussel and oyster beds, taken from one area. In the

next step the collected spectral signatures are used and extrapolated to a wider area (Figure 15).

Figure 15- Result of a mapping of habitat t ypes using the IMARES algorith m . Satellite image wit h detectedmussel/oyster beds (black and white picture). Left: step 1 relating spectral signatures of different types of

sediment and mussel/oyster beds in a selected area with known locations of mussel beds and sediment

composition. Right: step 2 extrapolation of the results from left picture to a wider area. Image resolution- 2meters.


36/45


Figures 16 and 17 shows images that can be used for mapping areas where changes in salt marshes

occur, by showing developments in the occurrence of pioneer vegetation, changes in areas with cliff

erosion, changes in vegetation coverage, etc. Regular flooding of salt marshes delivers sediment to the

system which results in vertical growth and vegetation change. This function of salt marshes is an

important ecosystem service which can be included in coastal defence strategies. Algorithms for high

resolution salt marsh zonation mapping still need to be developed.

Figure 16- Salt marshes along the coast (Peazemerlannen, Friesland). Levee breaks Breaches in dikes bordering

the former summer polders (1976) are highlighted in red circles (image date: October 2006). The flooded area

behind the broken dike level increased towards t he coast. In these areas salt m arsh rejuvenation occurs. Date:October 2 006

Figure 17- Salt m arsh zones (Westpolder, Groningen) in bright red (high salt m arsh), dark r ed (lower salt m arsh

and pioneer zone) and brown ( diatoms and/ or pioneer zone). Date: July 2011.


37/45


6 DiscussionRemote sensing can be a very useful and relatively inexpensive tool for detecting specific characteristics

of emerging tidal flats. Detailed detection of the properties of these characteristics requires satellite data

taken during cloud free weather and low tide. For each historical satellite mission, on average four

images per year are available for the specific parts of the Dutch Wadden Sea. Considering the limitations

on usability of historical satellite data mentioned in Section 3.6. the number of images which can be used

for trend analysis and any other related analysis of sediments and mussel and oyster beds, is limited.

The historical Landsat data showed different characteristics of intertidal sediments in terms of water

content and sand and mud content. The locations of mussel and oyster beds can be fairly well detected

on images acquired during low tide. However, due to tidal variation even over small distances, additional

algorithm calibrations are necessary to increase classification accuracy and need to be developed. In

order to cover the total Dutch Wadden Sea area, multiple satellite images are needed, each targetingspecific sections of the Wadden Sea within a rather narrow low tide window. This may offer considerable

challenges when attempting to analyse dynamic developments for the whole Dutch Wadden Sea based on

series of images from different years and from different months. The use of radar images that do not

require cloud free conditions may prove useful to add information to multi-spectral data, for instance to

enhance the spatial resolution of 30 meter Landsat images to 20 meter and 15 meter ASTER images to

12 meter. This enhances visibility and detection of mussel and oyster beds and will possibly improve the

sediments classification. Another contribution from radar images is provision of information on surface

texture and roughness, which can be used in verification of areas that are n

The use of Remote sensing as a monitoring tool for coastal defence issues in the Wadden Sea

Documents

Transcript of The use of Remote sensing as a monitoring tool for coastal defence issues in the Wadden Sea